FactoryWager Wiki

Tenant: ops-snapshot

Tenant ops-snapshot Runs bun lib/operations/snapshot-cron.ts --once — every */10 * * * * UTC via spine in-process Bun.cron complement Proof ops-snapshot-cron-v1 (claim in lib/harness/proof.ts) Catalog lib/harness/maintenance.ts · MAINTENANCE_RUNBOOKS

Refreshes portal/Pages artifacts:

public/registry/ops-summary.json (dashboard cards; embeds proofTaxonomy slice when taxonomy audit runs first)
public/registry/proof-taxonomy-audit.json (subsystem contracts + consistency — written before ops-summary)
public/registry/monitoring.json
public/registry/static.json
public/registry/dod-queue.json (+ bakes #dod-embed on /portal/dod/)
public/registry/toc-ops.json (+ bakes #toc-embed on /portal/toc/ · ops-summary toc slice)
public/registry/compliance-board.json (+ enhancements/shadow · portal embed · ops-summary compliance slice; disable with --no-compliance / OPS_SNAPSHOT_COMPLIANCE=0) — one board → many projections (ops-summary.compliance · monitoring.compliance · health complianceBoard · /portal/compliance/) · tenant compliance-portal.md
public/registry/limit-raises.json (multi-factor raise context · 48h · capture missing rows via exportLimitRaisesSnapshot) — portal /portal/limits/ · agent snapshot on Pages · tenant partner-limits.md
tenant slices public/registry/{factory,science,tennis}/registry.json (when thin)
@factorywager/bun-utils-test + routing-test proofs
prediction report (unless --no-report)
public/registry/release-features.json channel meta merge (prefer-artifact; disable with --no-channel-meta / OPS_SNAPSHOT_CHANNEL_META=0) · claim channel-meta-verification-v1

Populate demos (Pages “looks live”): bun run ops:snapshot:demo — see Flag matrix below.

TOC board: toc-ops.md · /portal/toc/ · concern matrix (rails/accounts/Soft/bots vs Cloudflare MCP) in that tenant.

Uses routing proof retry + TTL cache (lib/routing-proof.ts · getRoutingProof). Next fire waits for the snapshot Promise (no overlap).

Flag matrix (seed + compliance)

SSOT: tools/ops-snapshot.ts. Do not invent flags — only argv/Bun.env keys the tool actually reads.

Compliance board

Control	Default	Effect
(none)	on	Bake `compliance-board.json` (+ enhancements/shadow + portal embed) before ops-summary so `payload.compliance` / monitoring slice stay fresh
`--no-compliance`	—	Skip companion bake
`OPS_SNAPSHOT_COMPLIANCE=0`	—	Same skip (env; either flag or env disables)

bun run ops:snapshot -- --no-compliance
OPS_SNAPSHOT_COMPLIANCE=0 bun run ops:snapshot

Companion path: compliance-portal.md · bun run compliance:bake.

Seed flags

Auto-seed block (ops / prediction / DOD / partner bindings / tenant registries / TOC demo seed) runs unless --no-seed. Outside that block, snapshot always refreshes catalog-snapshot.json and re-exports toc-ops.json (+ soft/identity bridge).

Flag	When seed runs	`force`	`ifEmpty`	Notes
(default)	When the domain is empty (`is*Empty` / missing `toc-ops.json`)	`false`	`true`	No-op if data already present
`--seed`	Always enter each seed call	`false`	`false`	Bypasses the empty gate (`ifEmpty: false`) so seeders run even when data exists (inserts more demo rows; not a wipe)
`--seed-force`	Always	`true`	`false`	Overwrite / re-seed; also force-seeds TOC identity bridge + Soft Balance on the always-on TOC export
`--seed-tenants`	(with seed block)	tenants only: `true`	—	`seedTenantRegistries({ force: true })`; other domains keep default/seed/`--seed-force` rules
`--no-seed`	Never	—	—	Skips the whole demo-seed block (catalog + TOC export still run)

Combinations (as coded):

--seed-force implies seed want (wantSeed); sets forceSeed for all domains in the block.
--seed-tenants alone does not set wantSeed / forceSeed for ops/prediction/DOD/partners/TOC; it only ORs into tenant-registry force.
--no-seed wins: seed flags are ignored when the block is skipped.
Tenant force = forceSeed or --seed-tenants (so --seed-force also force-fills tenant registries).

bun run ops:snapshot -- --no-seed              # pure assemble; no demo fill
bun run ops:snapshot -- --seed                 # call seeders (empty-or-present policy per seeder)
bun run ops:snapshot -- --seed-force           # re-seed demos
bun run ops:snapshot -- --seed-tenants         # force thin tenant registry.json only

`ops:snapshot:demo`

bun run ops:snapshot:demo
# ≡  bun run ops:seed:all && bun tools/ops-snapshot.ts --no-routing

Step	What runs
1. `ops:seed:all`	External seeds: ops + prediction + DOD + partners + TOC fixture + tenants (`package.json` chain)
2. `ops-snapshot --no-routing`	Full snapshot without live routing proof (faster local/Pages “looks live”); seed block still default (if-empty) — usually no-ops after step 1

Compliance, channel-meta, static, report, TOC export, catalog stay on (defaults). ops:snapshot:demo does not forward extra argv; for a compliance-skip demo bake, run the two steps manually:

bun run ops:seed:all && bun tools/ops-snapshot.ts --no-routing --no-compliance

Other flags (reference)

Flag / env	Default	Effect
`--no-routing`	routing on	Skip `getRoutingProof`
`--force-routing`	cache OK	Force routing refresh
`--no-report`	report on	Skip prediction HTML/SVG
`--webview`	off	Prediction report WebView PNG
`--no-static`	static on	Skip `public/registry/static.json`
`--no-channel-meta` / `OPS_SNAPSHOT_CHANNEL_META=0`	channel meta on	Skip prefer-artifact channel meta merge
`--publish` / `OPS_SNAPSHOT_PUBLISH=1`	off	Multipart publish proof artifacts (needs `REGISTRY_URL` + secret)
`--out <path>` / `OPS_SNAPSHOT_PATH`	`public/registry/ops-summary.json`	Ops-summary write path
`OPS_DB_PATH`	default ops DB	SQLite path

Package scripts: ops:snapshot · ops:snapshot:demo · ops:snapshot:once (cron --once) · ops:snapshot:cron.

Signal (failure)

bun run spine:schedule:once -- --tenant=ops-snapshot exits non-zero. Also: missing/stale public/registry/ops-summary.json, or ops dashboard routing card shows criticalFailed > 0 after a successful write.

Proofs green but liquidity/plays empty → read ops-summary-endpoint.md (two pipelines; empty ops DB is not a failed summary assemble). Diagnose: bun run ops:diagnose.

Intervention (repair)

Catalog intervention: bun test tests/ops-snapshot-cron.test.ts · bun lib/operations/snapshot-cron.ts --once

Re-run: bun run spine:schedule:once -- --tenant=ops-snapshot
Local smoke: bun run ops:snapshot -- --no-routing
Force routing: bun lib/operations/snapshot-cron.ts --once --force-routing
Check data/operations.db and network reachability of the Pages public origin (ROUTING_PROBE_BASE_URL or default https://score.factory-wager.com). REGISTRY_URL is the npm registry API — not the routing probe target.

Do not delete the tenant to green the daemon.

Retirement

Remove when a required CI/Pages deploy pipeline always runs ops:snapshot before publish and spine is no longer the periodic owner. Keep spine-multi-tenant ≥2 via another tenant.

Retirement verified false — set tenants.ops-snapshot=true in lib/harness/ci-owned-tenants.json when CI owns the periodic refresh.

Retirement check ops-snapshot is owned by CI/Pages deploy pipeline → bun scripts/retirement-check-ci-owner.ts --tenant=ops-snapshot

Owner // owner: platform / ops portal Fresh-rerun bun run docs:tenant-ops-snapshot

This site is open source. Improve this page.